Logical Bayesian networks
Daan Fierens, Hendrik Blockeel, Jan Ramon, Maurice Bruynooghe · Lirias · 2004
Several models combining Bayesian networks with logic exist. The two most developed models are Probabilistic Relational Models (PRM's) and Bayesian Logic Programs (BLP's). While PRM's are easier to understand, BLP's are more expressive. However, we argue that BLP's do not always allow modeling problems intuitively. This motivates us to introduce Logical Bayesian Networks (LBN's). We argue that LBN's provide an expressive and intuitive modeling language due to explicitly distinguishing deterministic and probabilistic information and having multiple components. We briefly discuss perspectives for learning LBN's from data.